PASS1A (Rat) Metabolomics:
Mayo Clinic Replicate Comparisons (Pairs)

Setup the Environment

CSS Styling

writeLines("td, th { padding : 3px } th { background-color:white; color:black; border:1px solid black; text-align:center } td {color:black; border:1px solid black; word-wrap:break-word; white-space:nowrap; overflow: hidden; text-overflow: ellipsis; max-width:300px; text-align:left}", con= "../css/style.css")

Comparison of Samples and Metabolites Between Replicate Pairs

• Split data by replicates
• Determine the number of missing values per metabolite
• Determine the number of NA values per metabolite
• Plot boxplots and matched density plot before normalization
• Normalize
• Plot boxplots and matched density plot after normalization

Boxplots (Original Abundances)

# Plot Boxplots faceted by shared metabolites
################################################################################
p <- plot_df %>%
  ggplot(aes(y = METABOLITE_NAME, x = VALUE, color = REPLICATE)) +
  geom_boxplot(alpha = 0.8) +
  labs(title=paste0(tissue,' ',study_institute_metab_family,": Original Metabolite Abundances"),
             x = "Abundance", y = "") 
plot(p)

Zero/Missing Values in Metabolites (Original Abundances)

# Identify Zero/Missing Values in Metabolites
################################################################################
na_df <- plot_df %>%
  ungroup() %>% 
  group_by(METABOLITE_NAME, REPLICATE) %>%
  mutate(ZERO_LOG = ifelse(VALUE == 0, 1, 0)) %>%
  mutate(ZERO_N = sum(ZERO_LOG)) %>%
  mutate(ZERO_FREQ = round(ZERO_N/length(shared_sams), digits = 2)) %>%
  mutate(NA_LOG = ifelse(is.na(VALUE), 1, 0)) %>%
  mutate(NA_N = sum(NA_LOG)) %>%
  mutate(NA_FREQ = round(NA_N/length(shared_sams), digits = 2)) %>%
  select(REPLICATE, METABOLITE_NAME, ZERO_N, ZERO_FREQ, NA_N, NA_FREQ) %>%
  unique() %>%
  arrange(REPLICATE, desc(ZERO_FREQ), desc(NA_FREQ)) %>%
  ungroup()

na_df %>%
  knitr::kable(format = "html") %>%
  scroll_box(width = "100%", height = "400px")

REPLICATE	METABOLITE_NAME	ZERO_N	ZERO_FREQ	NA_N	NA_FREQ
811	Homocysteine	71	0.91	0	0
811	1-Methylhistidine	45	0.58	0	0
811	Carnosine	45	0.58	0	0
811	3-Methylhistidine	19	0.24	0	0
811	Arginine	4	0.05	0	0
811	Cysteine	1	0.01	0	0
811	Alanine	0	0.00	0	0
811	Anserine	0	0.00	0	0
811	Asparagine	0	0.00	0	0
811	Aspartic acid	0	0.00	0	0
811	Citrulline	0	0.00	0	0
811	Cystathionine	0	0.00	0	0
811	Ethanolamine	0	0.00	0	0
811	Glutamic acid	0	0.00	0	0
811	Glutamine	0	0.00	0	0
811	Glycine	0	0.00	0	0
811	Histidine	0	0.00	0	0
811	Hydroxyproline	0	0.00	0	0
811	Isoleucine	0	0.00	0	0
811	Leucine	0	0.00	0	0
811	Lysine	0	0.00	0	0
811	Methionine	0	0.00	0	0
811	Ornithine	0	0.00	0	0
811	Phenylalanine	0	0.00	0	0
811	Phosphoethanolamine	0	0.00	0	0
811	Proline	0	0.00	0	0
811	Sarcosine	0	0.00	0	0
811	Serine	0	0.00	0	0
811	Taurine	0	0.00	0	0
811	Threonine	0	0.00	0	0
811	Tryptophan	0	0.00	0	0
811	Tyrosine	0	0.00	0	0
811	Valine	0	0.00	0	0
811	2-Aminobutyric acid	0	0.00	0	0
811	Aminoadipic acid	0	0.00	0	0
811	beta-Alanine	0	0.00	0	0
811	Aminoisobutyric acid	0	0.00	0	0
811	gamma-Aminobutyric acid	0	0.00	0	0
812	Homocysteine	69	0.88	0	0
812	1-Methylhistidine	49	0.63	0	0
812	Carnosine	39	0.50	0	0
812	3-Methylhistidine	28	0.36	0	0
812	Arginine	2	0.03	0	0
812	Cysteine	1	0.01	0	0
812	Alanine	0	0.00	0	0
812	Anserine	0	0.00	0	0
812	Asparagine	0	0.00	0	0
812	Aspartic acid	0	0.00	0	0
812	Citrulline	0	0.00	0	0
812	Cystathionine	0	0.00	0	0
812	Ethanolamine	0	0.00	0	0
812	Glutamic acid	0	0.00	0	0
812	Glutamine	0	0.00	0	0
812	Glycine	0	0.00	0	0
812	Histidine	0	0.00	0	0
812	Hydroxyproline	0	0.00	0	0
812	Isoleucine	0	0.00	0	0
812	Leucine	0	0.00	0	0
812	Lysine	0	0.00	0	0
812	Methionine	0	0.00	0	0
812	Ornithine	0	0.00	0	0
812	Phenylalanine	0	0.00	0	0
812	Phosphoethanolamine	0	0.00	0	0
812	Proline	0	0.00	0	0
812	Sarcosine	0	0.00	0	0
812	Serine	0	0.00	0	0
812	Taurine	0	0.00	0	0
812	Threonine	0	0.00	0	0
812	Tryptophan	0	0.00	0	0
812	Tyrosine	0	0.00	0	0
812	Valine	0	0.00	0	0
812	2-Aminobutyric acid	0	0.00	0	0
812	Aminoadipic acid	0	0.00	0	0
812	beta-Alanine	0	0.00	0	0
812	Aminoisobutyric acid	0	0.00	0	0
812	gamma-Aminobutyric acid	0	0.00	0	0

# Remove Metabolites with high NAor Zero Frequencies
################################################################################
met_rm <- na_df %>%
  filter((ZERO_FREQ >= 0.95) | (NA_FREQ >= 0.8)) %>%
  select(METABOLITE_NAME) %>% unlist() %>% unique()
metabolites <- metabolites[metabolites %!in% met_rm]
plot_df <- plot_df %>%
  filter(METABOLITE_NAME %in% metabolites)
plot_df$METABOLITE_NAME <- as.character(plot_df$METABOLITE_NAME)

Distributions (Original Abundances)

# Summary Statistics of Density plot distribution
################################################################################
# Iterate through the runs and collect vectors
df_all <- data.frame()
for(rep in c(reps)){
  # Collect numeric vector
  num_vec <- plot_df %>%
    filter(REPLICATE == rep) %>%
    select(VALUE) %>% unlist() %>% unname()
  df <- NumericSummaryStats(num_vec)
  row.names(df) <- rep
  df_all <- rbind(df_all, df)
}
df_all %>%
  knitr::kable(format = "html") %>%
  scroll_box(width = "100%", height = "200px")

	NA_COUNT	NA_FREQ	MEAN	MEDIAN	MAX	MIN	MID_RANGE	VARIANCE	STD_DEV	SE_MEAN	Q1	Q3	KURTOSIS	SKEW	BC_LAMBDA
811	0	0	4.02	0.86	73.84	0	73.84	80.12	8.95	0.16	0.1	3.19	16.92	3.86	None
812	0	0	3.99	0.86	75.94	0	75.94	75.37	8.68	0.16	0.1	3.18	15.50	3.69	None

# Plot the density plot for all the gene counts
################################################################################
plot_df %>%
  ggplot(aes(x = VALUE, color = REPLICATE)) +
  geom_density() +
  xlim(min(df_all$MIN), mean(df_all$Q3)) +
  labs(title=paste0(tissue,' ',study_institute_metab_family,": Original Metabolite Abundances"),
       x = "Abundance",
       y = "Density")

Metabolite by Metabolite Correlation (Original Abundances)

################################################################################
####################### Metabolite by Metabolite Correlations ##################
################################################################################
# Subset Matrices
x <- plot_df %>%
  unite(METABOLITE_NAME,REPLICATE, 
        col = METABOLITE_NAME_REPLICATE, remove = F) %>%
  select(METABOLITE_NAME_REPLICATE, labelid, VALUE,REPLICATE) %>%
  split(.$REPLICATE) %>%
  map(select, -REPLICATE) %>%
  map(pivot_wider, names_from = METABOLITE_NAME_REPLICATE, values_from = VALUE) %>%
  map(arrange, labelid) %>%
  map(tibble::column_to_rownames, var = "labelid") %>%
  map(as.matrix)

# Subset matrices
if(all(row.names(x[[1]]) == row.names(x[[2]]))){
  shared_met_mat <-  do.call(cbind,x) 
}

Corr <- 'Spearman'
# Create an NxN Matrix of correlation
if(Corr == 'Pearson'){
        cor1 <- stats::cor(shared_met_mat, method = 'pearson') %>% round(digits = 3)
}else if(Corr == 'Spearman'){
        cor1 <- stats::cor(shared_met_mat, method = 'spearman') %>% round(digits = 3)
}


# Reorder the correlation matrix
cormat <- cor1

# Melt the correlation matrix
melted_cormat <- melt(cormat, na.rm = TRUE)

################################################################################
####################### Only Reciprocal Pairs ##################################
################################################################################

# Remove rows that compare metabolites from the same dataset, them remove redundent measurements
melted_cormat2 <- melted_cormat 
# Widen the datafrmae back to matrix
cormat2 <- melted_cormat2 %>%
  pivot_wider(names_from = Var2, values_from = value) %>%
  as.data.frame()
row.names(cormat2) <- cormat2$Var1
cormat2 <- cormat2 %>%
  select(-Var1) %>%
  as.matrix()
cormat2 <- cormat2[sort(rownames(cormat2)),sort(colnames(cormat2))]
cormat2 <- cormat2[grepl(reps[1], rownames(cormat2)),grepl(reps[2], colnames(cormat2))]

# Orient the colors and breaks
paletteLength <- 1000
myColor <- colorRampPalette(c("blue", "white", "red"))(paletteLength)
# length(breaks) == length(paletteLength) + 1
# use floor and ceiling to deal with even/odd length pallettelengths
myBreaks <- c(seq(-1, 0, length.out=ceiling(paletteLength/2) + 1), 
              seq(1/paletteLength, max(cor1, na.rm = T), length.out=floor(paletteLength/2)))

# Plot the heatmap
pheatmap(cormat2, color=myColor, breaks=myBreaks, 
                  cluster_rows = F, cluster_cols = F,fontsize = 6)

Scatterplots (Original Abundances)

# Subset the data for sites to compare (scatterplots)
###################################################
plot_join_df <- countdata_df %>%
  ungroup() %>%
  unite(STUDY_INSTITUTE,METAB_FAMILY, col = "STUDY_INSTITUTE_METAB_FAMILY") %>%
  filter(TISSUE == tissue) %>%
  filter(NAMED == named) %>%
  filter(STUDY_INSTITUTE_METAB_FAMILY == study_institute_metab_family) %>%
  unnest(COUNT_DATA) %>%
  filter(viallabel %in% as.character(unlist(sample_join$sample_id))) %>%
  filter(METABOLITE_NAME %in% metabolites) %>%
  unite(labelid, viallabel, col = "labelid_viallabel", remove = F) %>%
  mutate(REPLICATE = case_when(grepl(paste0(reps[1],"$"),labelid_viallabel) ~ reps[1],
                              grepl(paste0(reps[2],"$"),labelid_viallabel) ~ reps[2],
                              )) %>%
  select(REPLICATE,METABOLITE_NAME,labelid,VALUE) %>%
  mutate(REPLICATE = as.character(REPLICATE)) %>%
  split(.$REPLICATE) %>%
  map(~rename(., !!sym(unique(.$REPLICATE)) := "VALUE")) %>%
  map(~select(., -REPLICATE)) %>%
  purrr::reduce(left_join, by = c("labelid","METABOLITE_NAME"))

# Plot scatter plots
################################################################################
# Plot scatter plots ordered by sample (include R2 values)
lm_df <- plot_join_df %>%
  rename(name = METABOLITE_NAME) %>%
  rename(x = reps[1]) %>%
  rename(y = reps[2])
# Iterate through each of the shared metabolites to collect summary info about their correlations
r2_df <- data.frame()
for(metab in metabolites){
  met_df <- lm_df %>%
    filter(name == metab)
  r2_df <- rbind(r2_df, lm_eqn(met_df))
}
# Display summaries
met_r2_df <- r2_df %>%
  arrange(METABOLITE) %>%
  select(METABOLITE,R2) %>%
  arrange(desc(R2))
met_r2_df %>%
  knitr::kable(format = "html") %>%
  scroll_box(width = "100%", height = "200px")

METABOLITE	R2
2-Aminobutyric acid	0.913
Aminoadipic acid	0.894
Serine	0.728
Glutamine	0.706
Threonine	0.699
Aminoisobutyric acid	0.665
Alanine	0.654
Sarcosine	0.644
Lysine	0.638
Glutamic acid	0.605
Aspartic acid	0.597
Taurine	0.579
Glycine	0.568
Hydroxyproline	0.551
Cystathionine	0.489
3-Methylhistidine	0.487
Ornithine	0.459
Cysteine	0.305
Proline	0.298
Valine	0.277
Isoleucine	0.254
Tyrosine	0.246
Citrulline	0.23
beta-Alanine	0.207
Histidine	0.196
Tryptophan	0.173
gamma-Aminobutyric acid	0.146
Carnosine	0.121
Ethanolamine	0.121
Leucine	0.0946
Arginine	0.0774
Anserine	0.0758
Asparagine	0.0669
Methionine	0.0628
Phenylalanine	0.0324
Homocysteine	0.00266
1-Methylhistidine	0.000219
Phosphoethanolamine	0.000142

# Collect the order of metabolites
scat_met_order <- met_r2_df %>% select(METABOLITE) %>% unlist() %>% as.character()

# Plot the scatter plots
for(metab in scat_met_order){
  p <- plot_join_df %>%
    filter(METABOLITE_NAME == metab) %>%
  ggplot(aes(x = !!sym(reps[1]), y = !!sym(reps[2])), color = ) +
  geom_point(alpha = 0.5) +
  geom_smooth(method = "lm", size = 1) +
  geom_abline(linetype="dashed") +
  facet_wrap(~ METABOLITE_NAME) +
  expand_limits(x = 0, y = 0) +
  #labs(title=paste0(tissue,' ',study_institute_metab_family,": Normalized Metabolite Abundances")) +
  coord_fixed(ratio=1)
  plot(p)
}

Sample by Sample Correlations (Original Abundances)

################################################################################
####################### Sample by Sample Correlations ##########################
################################################################################

# Subset Data
################################################################################
# Original NxP Matrix
x <- plot_df %>%
  unite(labelid,REPLICATE, 
        col = labelid_REPLICATE, remove = F) %>%
  select(labelid_REPLICATE, METABOLITE_NAME, VALUE,REPLICATE) %>%
  split(.$REPLICATE) %>%
  map(select, -REPLICATE) %>%
  map(pivot_wider, names_from = METABOLITE_NAME, values_from = VALUE) %>%
  map(tibble::column_to_rownames, var = "labelid_REPLICATE") %>%
  map(as.matrix)
# Subset matrices
shared_sam_mat <-  do.call(rbind,x) %>% 
  t()

# Create an NxN Matrix of correlation
Corr <- 'Spearman'
if(Corr == 'Pearson'){
        cor1 <- stats::cor(shared_sam_mat, method = 'pearson') %>% round(digits = 3)
}else if(Corr == 'Spearman'){
        cor1 <- stats::cor(shared_sam_mat, method = 'spearman') %>% round(digits = 3)
}

# Melt the correlation matrix
melted_cormat <- melt(cor1, na.rm = TRUE)

################################################################################
####################### Unclustered ############################################
################################################################################

# Remove rows that compare metabolites from the same dataset, them remove redundent measurements
melted_cormat2 <- melted_cormat

# Widen the dataframwe back to matrix
cormat2 <- melted_cormat2 %>%
  pivot_wider(names_from = Var2, values_from = value) %>%
  as.data.frame()
row.names(cormat2) <- cormat2$Var1
cormat2 <- cormat2 %>%
  select(-Var1) %>%
  as.matrix()
cormat2 <- cormat2[sort(rownames(cormat2)),sort(colnames(cormat2))]
cormat2 <- cormat2[grepl(reps[1], rownames(cormat2)),grepl(reps[2], colnames(cormat2))]

# Orient the colors and breaks
paletteLength <- 1000
myColor <- colorRampPalette(c("blue", "white", "red"))(paletteLength)
# length(breaks) == length(paletteLength) + 1
# use floor and ceiling to deal with even/odd length pallettelengths
myBreaks <- c(seq(-1, 0, length.out=ceiling(paletteLength/2) + 1), 
              seq(1/paletteLength, max(cor1, na.rm = TRUE), length.out=floor(paletteLength/2)))

# Plot the heatmap
pheatmap(as.matrix(cormat2), color=myColor, breaks=myBreaks, 
                  cluster_rows = F, cluster_cols = F,fontsize = 6)

PCA Analysis (Original Abundances)

pca <- prcomp(na.omit(t(shared_sam_mat)), scale. = F)
percentVar <- pca$sdev^2/sum(pca$sdev^2)
PC <- pca$x %>% as.data.frame()
PC$labelid_rep <- as.character(row.names(pca$x))
PC <- PC %>%
  tidyr::separate(col = labelid_rep, into = c('labelid', 'REPLICATE'),
  sep = "_", remove = T)
# Join the phenotype data
PC <- left_join(PC, pheno_df, by = 'labelid')

p0 <- PC %>%
  ggplot(aes(x = PC1, y = PC2, color=REPLICATE)) +
  geom_point(size = 3) +
  geom_line(aes(group = labelid),color="dark grey") +
  ggtitle('Shared Samples and Shared Metabolites') + 
  xlab(paste0('PC1: ',round(percentVar[1]*100,2),'% variance')) +
  ylab(paste0('PC2: ',round(percentVar[2]*100,2),'% variance')) +
  theme(aspect.ratio=1)
p0

p1 <- PC %>%
        ggplot(aes(x = PC1, y = PC2, color=Key.anirandgroup, shape = REPLICATE)) +
        geom_point(size = 3) +
  geom_line(aes(group = labelid),color="dark grey") +
        ggtitle('Shared Samples and Shared Metabolites') + 
        xlab(paste0('PC1: ',round(percentVar[1]*100,2),'% variance')) +
        ylab(paste0('PC2: ',round(percentVar[2]*100,2),'% variance')) +
        scale_color_manual(values=ec_colors) +
        theme(aspect.ratio=1)
p1

PC$Registration.sex <- as.character(PC$Registration.sex)
p2 <- PC %>%
        ggplot(aes(x = PC1, y = PC2, color=Registration.sex, shape = REPLICATE)) +
        geom_point(size = 3) +
  geom_line(aes(group = labelid),color="dark grey") +
        ggtitle('Shared Samples and Shared Metabolites') + 
        xlab(paste0('PC1: ',round(percentVar[1]*100,2),'% variance')) +
        ylab(paste0('PC2: ',round(percentVar[2]*100,2),'% variance')) +
        theme(aspect.ratio=1)
p2

Normalization (AutoScale)

# Normalize
# Note in this step, we use labelid to account for mayo's duplicate samples
################################################################################
norm_df <- plot_df %>%
  unite(labelid, viallabel, col = "labelid_viallabel", remove = F) %>%
  split(.$REPLICATE) %>%
  map(select, labelid_viallabel, METABOLITE_NAME, VALUE) %>%
  map(pivot_wider, names_from = METABOLITE_NAME, values_from = VALUE) %>%
  map(column_to_rownames, var = "labelid_viallabel") %>%
  map(as.matrix) %>%
  map(AutoScaleMatrix) %>%
  map(as.data.frame) %>%
  map(rownames_to_column, var = "labelid_viallabel") %>%
  map(pivot_longer, names_to = 'METABOLITE_NAME', values_to = 'VALUE', cols = all_of(metabolites)) %>%
  map(mutate, REPLICATE = case_when(grepl(paste0(reps[1],"$"),labelid_viallabel) ~ reps[1],
                                     grepl(paste0(reps[2],"$"),labelid_viallabel) ~ reps[2],
                                    grepl(paste0(reps[3],"$"),labelid_viallabel) ~ reps[3])) %>%
  bind_rows()

Boxplots (Normalized Abundances)

# Plot Boxplots faceted by shared metabolites
################################################################################
norm_df %>%
  ggplot(aes(y = METABOLITE_NAME, x = VALUE, color = REPLICATE)) +
  geom_boxplot(alpha = 0.8) +
  labs(title=paste0(tissue,' ',study_institute_metab_family,": Normalized Metabolite Abundances"),
             x = "Abundance", y = "") 

Metabolite by Metabolite Correlation (Normalized Abundances)

################################################################################
####################### Metabolite by Metabolite Correlations ##################
################################################################################
# Subset Matrices
x <- plot_df %>%
  unite(METABOLITE_NAME,REPLICATE, 
        col = METABOLITE_NAME_REPLICATE, remove = F) %>%
  select(METABOLITE_NAME_REPLICATE, labelid, VALUE,REPLICATE) %>%
  split(.$REPLICATE) %>%
  map(select, -REPLICATE) %>%
  map(pivot_wider, names_from = METABOLITE_NAME_REPLICATE, values_from = VALUE) %>%
  map(arrange, labelid) %>%
  map(tibble::column_to_rownames, var = "labelid") %>%
  map(as.matrix) %>%
  map(AutoScaleMatrix)

# Subset matrices
if(all(row.names(x[[1]]) == row.names(x[[2]]))){
  shared_met_mat <-  do.call(cbind,x) 
}

Corr <- 'Spearman'
# Create an NxN Matrix of correlation
if(Corr == 'Pearson'){
        cor1 <- stats::cor(shared_met_mat, method = 'pearson') %>% round(digits = 3)
}else if(Corr == 'Spearman'){
        cor1 <- stats::cor(shared_met_mat, method = 'spearman') %>% round(digits = 3)
}

# Reorder the correlation matrix
cormat <- cor1

# Melt the correlation matrix
melted_cormat <- melt(cormat, na.rm = TRUE)

################################################################################
####################### Only Reciprocal Pairs ##################################
################################################################################

# Remove rows that compare metabolites from the same dataset, them remove redundent measurements
melted_cormat2 <- melted_cormat 
# Widen the datafrmae back to matrix
cormat2 <- melted_cormat2 %>%
  pivot_wider(names_from = Var2, values_from = value) %>%
  as.data.frame()
row.names(cormat2) <- cormat2$Var1
cormat2 <- cormat2 %>%
  select(-Var1) %>%
  as.matrix()
cormat2 <- cormat2[sort(rownames(cormat2)),sort(colnames(cormat2))]
cormat2 <- cormat2[grepl(reps[1], rownames(cormat2)),grepl(reps[2], colnames(cormat2))]

# Orient the colors and breaks
paletteLength <- 1000
myColor <- colorRampPalette(c("blue", "white", "red"))(paletteLength)
# length(breaks) == length(paletteLength) + 1
# use floor and ceiling to deal with even/odd length pallettelengths
myBreaks <- c(seq(-1, 0, length.out=ceiling(paletteLength/2) + 1), 
              seq(1/paletteLength, max(cor1, na.rm = T), length.out=floor(paletteLength/2)))

# Plot the heatmap
pheatmap(cormat2, color=myColor, breaks=myBreaks, 
                  cluster_rows = F, cluster_cols = F,fontsize = 6)

Scatterplots (Normalized Abundances)

# Normalize
# Subset the data for sites to compare (scatterplots)
###################################################
norm_join_df <- countdata_df %>%
  ungroup() %>%
  unite(STUDY_INSTITUTE,METAB_FAMILY, col = "STUDY_INSTITUTE_METAB_FAMILY") %>%
  filter(TISSUE == tissue) %>%
  filter(NAMED == named) %>%
  filter(STUDY_INSTITUTE_METAB_FAMILY == study_institute_metab_family) %>%
  unnest(COUNT_DATA) %>%
  filter(viallabel %in% as.character(unlist(sample_join$sample_id))) %>%
  unite(labelid, viallabel, col = "labelid_viallabel", remove = F) %>%
  mutate(REPLICATE = case_when(grepl(paste0(reps[1],"$"),labelid_viallabel) ~ reps[1],
                                     grepl(paste0(reps[2],"$"),labelid_viallabel) ~ reps[2],
                               grepl(paste0(reps[3],"$"),labelid_viallabel) ~ reps[3])) %>%
  split(.$REPLICATE) %>%
  map(select, labelid_viallabel, METABOLITE_NAME, VALUE) %>%
  map(pivot_wider, names_from = METABOLITE_NAME, values_from = VALUE) %>%
  map(column_to_rownames, var = "labelid_viallabel") %>%
  map(as.matrix) %>%
  map(AutoScaleMatrix) %>%
  map(as.data.frame) %>%
  map(rownames_to_column, var = "labelid_viallabel") %>%
  map(pivot_longer, names_to = 'METABOLITE_NAME', values_to = 'VALUE', cols = all_of(metabolites)) %>%
  map(mutate, REPLICATE = case_when(grepl(paste0(reps[1],"$"),labelid_viallabel) ~ reps[1],
                                     grepl(paste0(reps[2],"$"),labelid_viallabel) ~ reps[2],
                                    grepl(paste0(reps[3],"$"),labelid_viallabel) ~ reps[3])) %>%
  map(~rename(., !!sym(unique(.$REPLICATE)) := "VALUE")) %>%
  map(~select(., -REPLICATE)) %>%
  map(mutate, labelid = gsub(pattern = "_..*", replacement = '', labelid_viallabel)) %>%
  map(select, -labelid_viallabel) %>%
  purrr::reduce(left_join, by = c("labelid","METABOLITE_NAME"))

# Plot scatter plots ordered by sample (include R2 values)
lm_df <- norm_join_df %>%
  rename(name = METABOLITE_NAME) %>%
  rename(x = reps[1]) %>%
  rename(y = reps[2])
# Iterate through each of the shared metabolites to collect summary info about their correlations
r2_df <- data.frame()
for(metab in metabolites){
  met_df <- lm_df %>%
    filter(name == metab)
  r2_df <- rbind(r2_df, lm_eqn(met_df))
}
# Display summaries
met_r2_df <- r2_df %>%
  arrange(METABOLITE) %>%
  select(METABOLITE,R2) %>%
  arrange(desc(R2))
met_r2_df %>%
  knitr::kable(format = "html") %>%
  scroll_box(width = "100%", height = "200px")

METABOLITE	R2
2-Aminobutyric acid	0.913
Aminoadipic acid	0.894
Serine	0.728
Glutamine	0.706
Threonine	0.699
Aminoisobutyric acid	0.665
Alanine	0.654
Sarcosine	0.644
Lysine	0.638
Glutamic acid	0.605
Aspartic acid	0.597
Taurine	0.579
Glycine	0.568
Hydroxyproline	0.551
Cystathionine	0.489
3-Methylhistidine	0.487
Ornithine	0.459
Cysteine	0.305
Proline	0.298
Valine	0.277
Isoleucine	0.254
Tyrosine	0.246
Citrulline	0.23
beta-Alanine	0.207
Histidine	0.196
Tryptophan	0.173
gamma-Aminobutyric acid	0.146
Carnosine	0.121
Ethanolamine	0.121
Leucine	0.0946
Arginine	0.0774
Anserine	0.0758
Asparagine	0.0669
Methionine	0.0628
Phenylalanine	0.0324
Homocysteine	0.00266
1-Methylhistidine	0.000219
Phosphoethanolamine	0.000142

# Collect the order of metabolites
scat_met_order <- met_r2_df %>% select(METABOLITE) %>% unlist() %>% as.character()

# Plot scatter plots (Normalized)
################################################################################
# Plot the scatter plots
for(metab in scat_met_order){
  p <- norm_join_df %>%
    filter(METABOLITE_NAME == metab) %>%
  ggplot(aes(x = !!sym(reps[1]), y = !!sym(reps[2])), color = ) +
  geom_point(alpha = 0.5) +
  geom_smooth(method = "lm", size = 1) +
  geom_abline(linetype="dashed") +
  facet_wrap(~ METABOLITE_NAME) +
  expand_limits(x = 0, y = 0) +
  #labs(title=paste0(tissue,' ',study_institute_metab_family,": Normalized Metabolite Abundances")) +
  coord_fixed(ratio=1)
  plot(p)
}

# norm_join_df %>%
#   mutate(METABOLITE_NAME = factor(METABOLITE_NAME, levels = scat_met_order)) %>%
#   ggplot(aes(x = !!sym(reps[1]), y = !!sym(reps[2])), color = ) +
#   geom_point(alpha = 0.5) +
#   geom_smooth(method = "lm", size = 1) +
#   geom_abline(linetype="dashed") +
#   facet_wrap(~ METABOLITE_NAME) +
#   expand_limits(x = 0, y = 0) +
#   labs(title=paste0(tissue,' ',study_institute_metab_family,": Normalized Metabolite Abundances")) +
#   coord_fixed(ratio=1)

Distributions (Normalized Abundances)

# Plot the density plot for all the gene counts
################################################################################
norm_df %>%
  ggplot(aes(x = VALUE, color = REPLICATE)) +
  geom_density() +
  labs(title=paste0(tissue,' ',study_institute_metab_family,": Normalized Metabolite Abundances"),
       x = "Abundance",
       y = "Density")

# Summary Statistics of Density plot distribution
################################################################################
# Iterate through the runs and collect vectors
df_all <- data.frame()
for(rep in reps){
  # Collect numeric vector
  num_vec <- norm_df %>%
    filter(REPLICATE == rep) %>%
    select(VALUE) %>% unlist() %>% unname()
  df <- NumericSummaryStats(num_vec)
  row.names(df) <- rep
  df_all <- rbind(df_all, df)
}
df_all
#}

Sample by Sample Correlations (Normalized Abundances)

################################################################################
####################### Sample by Sample Correlations ##########################
################################################################################

# Subset Data
################################################################################
# Original NxP Matrix
x <- plot_df %>%
  unite(labelid,REPLICATE, 
        col = labelid_REPLICATE, remove = F) %>%
  select(labelid_REPLICATE, METABOLITE_NAME, VALUE,REPLICATE) %>%
  split(.$REPLICATE) %>%
  map(select, -REPLICATE) %>%
  map(pivot_wider, names_from = METABOLITE_NAME, values_from = VALUE) %>%
  map(tibble::column_to_rownames, var = "labelid_REPLICATE") %>%
  map(as.matrix) %>%
  map(AutoScaleMatrix)
# Subset matrices
shared_sam_mat <-  do.call(rbind,x) %>% 
  t()

# Create an NxN Matrix of correlation
Corr <- 'Spearman'
if(Corr == 'Pearson'){
        cor1 <- stats::cor(shared_sam_mat, method = 'pearson') %>% round(digits = 3)
}else if(Corr == 'Spearman'){
        cor1 <- stats::cor(shared_sam_mat, method = 'spearman') %>% round(digits = 3)
}

# Melt the correlation matrix
melted_cormat <- melt(cor1, na.rm = TRUE)

################################################################################
####################### Unclustered ############################################
################################################################################

# Remove rows that compare metabolites from the same dataset, them remove redundent measurements
melted_cormat2 <- melted_cormat

# Widen the dataframwe back to matrix
cormat2 <- melted_cormat2 %>%
  pivot_wider(names_from = Var2, values_from = value) %>%
  as.data.frame()
row.names(cormat2) <- cormat2$Var1
cormat2 <- cormat2 %>%
  select(-Var1) %>%
  as.matrix()
cormat2 <- cormat2[sort(rownames(cormat2)),sort(colnames(cormat2))]
cormat2 <- cormat2[grepl(reps[1], rownames(cormat2)),grepl(reps[2], colnames(cormat2))]

# Orient the colors and breaks
paletteLength <- 1000
myColor <- colorRampPalette(c("blue", "white", "red"))(paletteLength)
# length(breaks) == length(paletteLength) + 1
# use floor and ceiling to deal with even/odd length pallettelengths
myBreaks <- c(seq(-1, 0, length.out=ceiling(paletteLength/2) + 1), 
              seq(1/paletteLength, max(cor1, na.rm = TRUE), length.out=floor(paletteLength/2)))

# Plot the heatmap
pheatmap(as.matrix(cormat2), color=myColor, breaks=myBreaks, 
                  cluster_rows = F, cluster_cols = F,fontsize = 6)

PCA Analysis (Normalized Abundances)

# Plot a PCA with outliers labeled
# shared_sam_mat <- shared_sam_mat %>%
#   t() %>% AutoScaleMatrix() %>%
#   t()
pca <- prcomp(na.omit(t(shared_sam_mat)), scale. = F)
percentVar <- pca$sdev^2/sum(pca$sdev^2)
PC <- pca$x %>% as.data.frame()
PC$labelid_rep <- as.character(row.names(pca$x))
PC <- PC %>%
  tidyr::separate(col = labelid_rep, into = c('labelid', 'REPLICATE'),
  sep = "_", remove = T)
# Join the phenotype data
PC <- left_join(PC, pheno_df, by = 'labelid')

p0 <- PC %>%
  ggplot(aes(x = PC1, y = PC2, color=REPLICATE)) +
  geom_point(size = 3) +
  geom_line(aes(group = labelid),color="dark grey") +
  ggtitle('Shared Samples and Shared Metabolites') + 
  xlab(paste0('PC1: ',round(percentVar[1]*100,2),'% variance')) +
  ylab(paste0('PC2: ',round(percentVar[2]*100,2),'% variance')) +
  theme(aspect.ratio=1)
p0

p1 <- PC %>%
        ggplot(aes(x = PC1, y = PC2, color=Key.anirandgroup, shape = REPLICATE)) +
        geom_point(size = 3) +
  geom_line(aes(group = labelid),color="dark grey") +
        ggtitle('Shared Samples and Shared Metabolites') + 
        xlab(paste0('PC1: ',round(percentVar[1]*100,2),'% variance')) +
        ylab(paste0('PC2: ',round(percentVar[2]*100,2),'% variance')) +
        scale_color_manual(values=ec_colors) +
        theme(aspect.ratio=1)
p1

PC$Registration.sex <- as.character(PC$Registration.sex)
p2 <- PC %>%
        ggplot(aes(x = PC1, y = PC2, color=Registration.sex, shape = REPLICATE)) +
        geom_point(size = 3) +
  geom_line(aes(group = labelid),color="dark grey") +
        ggtitle('Shared Samples and Shared Metabolites') + 
        xlab(paste0('PC1: ',round(percentVar[1]*100,2),'% variance')) +
        ylab(paste0('PC2: ',round(percentVar[2]*100,2),'% variance')) +
        theme(aspect.ratio=1)
p2